J. Mundy first gives a general overview of the domain of computer vision. Then a short introduction into affine geometries is given afterwards.
The vision algorithm, which he is proposing does the following steps (as far as I've understood):
*Extract edges
*Compare projected "vertex-pairs" with vertex pairs of the model
*Compute the corresponding transformations for each pair (configuration) to map the model in the scene
*Search for clusters in the configuration space.
*Find middle of cluster using least-squares. The center of the cluster gives the objects position in the scene.
The lecture is motivating and the results are quite impressive, exspecially if you consider, that it was presented in 1987.
J. Mundy also gives some advice, how a student should prepare himself for doing computer vision ;-)